Pressure testing method and device, storage medium and electronic device

Abstract

The embodiment of the invention provides a pressure testing method and device, a storage medium and an electronic device.The method comprises the steps that a historical yield fluctuation sequence of each target bond held by a tested target is utilized, parameters of an edge distribution function corresponding to each target bond are determined, and the multiple target bonds are divided into multiple levels; calculating model parameters of the Copula model corresponding to each hierarchy, and determining a multi-dimensional joint distribution function corresponding to the target bond in each hierarchy according to the model parameters of the Copula model corresponding to each hierarchy and the parameters of the edge distribution function corresponding to the target bond in each hierarchy; based on the multi-dimensional joint distribution function corresponding to the target bond in each hierarchy, generating a group of analog number pairs corresponding to yield fluctuation for the target bond in each hierarchy, splicing the group of analog number pairs generated for each hierarchy, and performing pressure test on the tested target by using at least part of the pressure scene obtained by splicing.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of the Internet, and specifically, to a stress testing method and device, a storage medium, and an electronic device. Background Art

[0002] In the relevant technology, the stress test of the bond market usually adopts the hypothetical scenario method based on expert experience. The hypothetical scenario method based on expert experience relies on the knowledge and experience of industry experts to construct the stress test scenario (for example, when analyzing the solvency sensitivity of the banking industry, this method can be used to set the stress scenario of investment loss risk): first, the bond holdings of the tested institution are divided into multiple types of bonds; then, according to expert experience, different yield curve upward shifts are set for each type of bond; finally, under different stress scenarios, assuming that the asset-liability structure of the bank remains unchanged, calculate whether the core tier 1 capital adequacy ratio, tier 1 capital adequacy ratio and capital adequacy ratio of the participating bank after the impact meet the regulatory requirements.

[0003] However, the hypothetical scenario method based on expert experience relies on experience and cognition, and the judgment process is relatively subjective, resulting in poor accuracy of the test results. Therefore, it can be seen that the stress testing method in the related art has the problem of poor accuracy of the test results. Summary of the invention

[0004] Embodiments of the present application provide a stress testing method and device, a storage medium, and an electronic device to at least solve the problem of poor accuracy of test results in stress testing methods in related arts.

[0005] According to one aspect of an embodiment of the present application, a stress testing method is provided, comprising: determining a historical yield fluctuation sequence of each target bond among a plurality of target bonds held by a target under test, wherein the plurality of target bonds are divided into a plurality of levels, each level in the plurality of levels contains at least two target bonds among the plurality of target bonds, and each level corresponds to a Copula model; using the historical yield fluctuation sequence of each target bond, determining the parameters of a marginal distribution function corresponding to each target bond; and calculating the model parameters of the Copula model corresponding to each level based on the historical yield fluctuation sequence of the target bonds in each level. , and determine the multidimensional joint distribution function corresponding to the target bonds in each level according to the model parameters of the Copula model corresponding to each level and the parameters of the marginal distribution function corresponding to the target bonds in each level; based on the multidimensional joint distribution function corresponding to the target bonds in each level, generate a set of simulated number pairs of corresponding yield fluctuations for the target bonds in each level, and splice the set of simulated number pairs generated for each level to obtain a set of stress scenarios for the target under test; use at least part of the stress scenarios in the set of stress scenarios to perform stress testing on the target under test to obtain the stress test results of the target under test.

[0006] According to another aspect of an embodiment of the present application, a stress testing device is also provided, including: a first determination unit, used to determine the historical yield fluctuation sequence of each target bond among a plurality of target bonds held by a target under test, wherein the plurality of target bonds are divided into a plurality of levels, each level of the plurality of levels contains at least two target bonds among the plurality of target bonds, and each level corresponds to a Copula model; a second determination unit, used to determine the parameters of the marginal distribution function corresponding to each target bond by using the historical yield fluctuation sequence of each target bond; a first execution unit, used to calculate the Copula model corresponding to each level based on the historical yield fluctuation sequence of the target bonds in each level The method comprises the following steps: a first executing unit, a second executing unit, and a second executing unit, for generating a set of simulated number pairs of yield fluctuations for the target bonds in each level based on the model parameters of the Copula model corresponding to each level and the parameters of the marginal distribution function corresponding to the target bonds in each level. The first executing unit is used to generate a set of simulated number pairs of yield fluctuations for the target bonds in each level based on the multidimensional joint distribution function corresponding to the target bonds in each level, and to splice the set of simulated number pairs generated for each level to obtain a set of stress scenarios for the target under test. The second executing unit is used to perform stress testing on the target under test using at least part of the stress scenarios in the set of stress scenarios to obtain stress testing results of the target under test.

[0007] According to another aspect of the embodiments of the present application, a computer-readable storage medium is provided, in which a computer program is stored. When the computer program is executed by a processor, the steps in any of the above method embodiments are implemented.

[0008] According to another aspect of the embodiments of the present application, an electronic device is also provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of any of the above method embodiments when executing the computer program.

[0009] According to another aspect of the embodiments of the present application, a computer program product or a computer program is also provided, the computer program product or the computer program includes computer instructions, and the computer instructions are stored in a computer-readable storage medium. The processor of the computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device performs the steps in any of the above method embodiments.

[0010] In an embodiment of the present application, a plurality of target bonds are divided into a plurality of levels, and a Copula model is applied at each level to capture the dependency between the yield fluctuations of the plurality of target bonds, so as to determine the historical yield fluctuation sequence of each target bond among the plurality of target bonds held by the measured target, wherein the plurality of target bonds are divided into a plurality of levels, each level in the plurality of levels contains at least two target bonds among the plurality of target bonds, and each level corresponds to a Copula model; the parameters of the marginal distribution function corresponding to each target bond are determined by using the historical yield fluctuation sequence of each target bond; based on the historical yield fluctuation sequence of the target bonds in each level, the model parameters of the Copula model corresponding to each level are calculated, and according to the model parameters of the Copula model corresponding to each level and the parameters of the marginal distribution function corresponding to the target bonds in each level, the marginal distribution function corresponding to each target bond is determined. A multidimensional joint distribution function corresponding to the target bonds in the hierarchy; based on the multidimensional joint distribution function corresponding to the target bonds in each hierarchy, a set of simulated number pairs of corresponding yield fluctuations are generated for the target bonds in each hierarchy respectively, and a set of simulated number pairs generated for each hierarchy are spliced ​​to obtain a set of stress scenarios for the target under test; at least part of the stress scenarios in a set of stress scenarios are used to perform stress testing on the target under test to obtain the stress test results of the target under test. Since multi-layer Copula technology is applied in the stress test to characterize the nonlinear correlation between the yield fluctuations of multiple target bonds and the over-fitting phenomenon of a single-layer Copula can be prevented, complex stress scenario modeling can be realized more accurately and scientifically, and the rationality of stress scenario setting is improved, thereby achieving the technical effect of improving the accuracy of the test results, thereby solving the problem of poor accuracy of the test results of the stress testing method in the related art. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 is a hardware structure block diagram of an optional terminal device according to an embodiment of the present application;

[0012] Figure 2 is a flow chart of an optional pressure testing method according to an embodiment of the present application;

[0013] Figure 3 is a schematic diagram of an optional pressure testing method according to an embodiment of the present application;

[0014] Figure 4 is a flow chart of another optional pressure testing method according to an embodiment of the present application;

[0015] Figure 5 is a structural block diagram of an optional pressure testing device according to an embodiment of the present application;

[0016] Figure 6 It is a block diagram of a computer system structure of an optional electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0017] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present application.

[0018] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0019] The method embodiments provided in the embodiments of the present application can be executed in a terminal device, a computer terminal or a similar computing device. Taking running on a terminal device as an example, Figure 1 is a hardware structure block diagram of an optional terminal device according to an embodiment of the present application. Figure 1 As shown, the terminal device may include one or more ( Figure 1 Only one is shown in the figure) a processor 102 (the processor 102 may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA) and a memory 104 for storing data, wherein the terminal device may also include a transmission device 106 and an input / output device 108 for communication functions. It can be understood by those skilled in the art that Figure 1 The structure shown is only for illustration and does not limit the structure of the above terminal device. Figure 1 More or fewer components as shown, or with Figure 1 Different configurations are shown.

[0020] The memory 104 can be used to store computer programs, for example, software programs and modules of application software, such as the computer program corresponding to the stress testing method in the embodiment of the present application. The processor 102 executes various functional applications and data processing by running the computer program stored in the memory 104, that is, to implement the above method. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include a memory remotely arranged relative to the processor 102, and these remote memories may be connected to the terminal device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0021] The transmission device 106 is used to receive or send data via a network. The specific example of the above network may include a wireless network provided by a communication provider of the terminal device. In one example, the transmission device 106 includes a network adapter (Network Interface Controller, referred to as NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device 106 can be a radio frequency (Radio Frequency, referred to as RF) module, which is used to communicate with the Internet wirelessly.

[0022] In this embodiment, a stress test method running on the above terminal device is provided. Figure 2 is a flow chart of an optional pressure testing method according to an embodiment of the present application, such as Figure 2 As shown, the process includes the following steps:

[0023] Step S202, determining a historical yield fluctuation sequence of each target bond among a plurality of target bonds held by the target under test, wherein the plurality of target bonds are divided into a plurality of levels, each level in the plurality of levels contains at least two target bonds among the plurality of target bonds, and each level corresponds to a Copula model;

[0024] Step S204, using the historical yield fluctuation sequence of each target bond, determining the parameters of the marginal distribution function corresponding to each target bond;

[0025] Step S206, based on the historical yield fluctuation sequence of the target bonds in each level, the model parameters of the Copula model corresponding to each level are calculated, and according to the model parameters of the Copula model corresponding to each level and the parameters of the marginal distribution function corresponding to the target bonds in each level, the multidimensional joint distribution function corresponding to the target bonds in each level is determined;

[0026] Step S208, based on the multidimensional joint distribution function corresponding to the target bonds in each level, a set of simulated number pairs of yield fluctuations corresponding to the target bonds in each level are generated respectively, and the set of simulated number pairs generated for each level are spliced ​​to obtain a set of stress scenarios for the measured target;

[0027] Step S210: perform a stress test on the target under test using at least part of the stress scenarios in a set of stress scenarios to obtain a stress test result of the target under test.

[0028] The stress testing method provided in this embodiment can be applied to the scenario of stress testing the target under test. Stress testing is a risk management and regulatory analysis tool used to analyze the impact of hypothetical, extreme but possible adverse scenarios on the financial institution as a whole, asset portfolio or other stress-sensitive targets under test (testing investment loss risk stress scenarios), and then evaluate its negative impact on the asset quality, profitability, capital level and liquidity of the institution. Stress testing is a conventional risk management tool, which is usually organized and carried out regularly by regulatory authorities, and each financial institution formulates an internal stress testing system according to regulatory requirements.

[0029] The process of stress testing the target usually includes determining the test target, determining the risk factors, setting the stress scenario, collecting test data, setting the assumptions, determining the test method, conducting the stress test, and analyzing the test results. After determining the test target and risk factors, setting the stress scenario is the starting point and foundation of the entire test process and is a key step to ensure that the stress test can be carried out effectively.

[0030] The stress testing method used in the relevant technology is a hypothetical scenario method based on expert experience. The hypothetical scenario method based on expert experience relies on the knowledge and experience of industry experts to construct the stress test scenario (for example, when analyzing the solvency sensitivity of the banking industry, this method can be used to set the stress scenario of investment loss risk): first, the bond holdings of the tested institution are divided into multiple types of bonds, such as interest-bearing bonds, commercial financial bonds and interbank certificates of deposit, and non-financial corporate bonds; then, according to expert experience, different yield curve upward shifts are set for each type of bond (as shown in Table 1); finally, under different stress scenarios, assuming that the bank's asset-liability structure remains unchanged, calculate whether the core tier 1 capital adequacy ratio, tier 1 capital adequacy ratio and capital adequacy ratio of the participating banks after the impact meet regulatory requirements.

[0031] Table 1 Banking solvency sensitivity stress scenarios

[0032]

[0033] However, the hypothetical scenario method based on expert experience relies on experience and cognition, the judgment process is relatively subjective, and lacks support from historical experience data and scientific quantitative models, resulting in poor accuracy of test results.

[0034] In order to make up for the shortcomings of the above methods, the historical scenario method based on historical percentiles can be used. The historical scenario method based on percentiles uses historical real data to determine the stress test scenario and determines the risk level under extreme conditions by analyzing historical data. It is usually assumed that different types of bonds are independent of each other, and the values ​​of their respective historical percentiles are selected as stress scenarios. For example, based on the historical data of the 1-year yield of treasury bonds and short-term notes of various grades, the annual yield upward range of 80%, 90% and 99% percentiles in the past five years is used as mild, moderate and severe stress scenarios to analyze the fair value loss of bond holdings under stress scenarios. In addition, when conducting regular stress tests on wealth management products, the historical percentiles of 90%, 95% and 99% of the yields of various types of bonds are taken as mild, moderate and severe stress scenarios to test the volatility of wealth management product returns.

[0035] However, the historical scenario method based on percentiles does not consider the correlation between the yields of different types of bonds. At the same time, the scenario set by this method needs to have actually occurred in history, which may underestimate the potential risks in the future. In addition, neither the hypothetical scenario method based on expert experience nor the historical scenario method based on percentiles considers the process of market dynamic changes under market risk scenarios.

[0036] It should be noted that, judging from the actual situation of the bond market, there are often complex linkages between the yields of various bonds. The fluctuation of bond yields is affected by macro factors, fundamental factors, market liquidity factors, emotional factors, etc. Different types of bonds are affected by various factors to different degrees, so their yields show asymmetric correlation characteristics. For example, when investors' risk appetite declines, while lowering various yields, they will relatively reduce the holdings of high-risk bonds, causing credit spreads to widen. Currently, whether it is the hypothetical scenario method or the historical scenario method, either independent yield changes are set for different types of bonds, or all types of bond yields are assumed to change in parallel. The former ignores the correlation between different yields, and the latter ignores the asymmetry of changes in different yields.

[0037] In addition, large fluctuations in bond yields often do not occur all at once, but rather present a dynamic change process over a certain period of time. The stress testing schemes in related technologies generally assume that external conditions experience rapid and extreme shocks, without considering the speed and frequency of market reactions. If the dynamic market changes are taken into account, the real impact faced by institutions can be simulated more accurately.

[0038] In view of the above-mentioned characteristics of the bond market, in this embodiment, a method for setting up a stress test scenario is provided to improve scientificity and effectiveness: a statistical model is used to model the volatility characteristics of various bond yields, capture their dynamic change characteristics, and generate a more realistic test scenario. The stress testing method in the related art cannot set up stress test scenarios for multiple types of yield curves based on their correlation. The setting of stress test scenarios in the bond market is actually to determine scenarios that conform to the law of yield fluctuations, which involves the linkage relationship of multiple types of bonds. From a statistical point of view, it is a problem of multivariate correlation processing. In this embodiment, a statistical model is used, combined with historical data on bond yield fluctuations, and the correlation of multi-dimensional bond fluctuations is considered to realize the automatic generation of bond market stress test scenarios, thereby optimizing the existing stress test scenario setting method.

[0039] In this embodiment, in order to set a stress scenario, the historical yield fluctuation sequence of each target bond among the multiple target bonds held by the target under test can be first determined. Here, the multiple target bonds are the bonds used for stress testing the target under test, which can be specified based on experience or selected from preset candidate bonds according to set screening conditions. The multiple target bonds are divided into multiple levels, each of the multiple levels contains at least two target bonds among the multiple target bonds, and each level corresponds to a Copula (connection) model.

[0040] In dealing with the problem of multivariate correlation, models such as correlation coefficient, multivariate linear regression, multivariate variance analysis, generalized linear model, multivariate time series, etc. often assume that the relationship between variables is linear and satisfies the normal distribution. For the bond market stress test scenario setting, it is assumed that the benchmark interest rate fluctuation satisfies the normal distribution, which is difficult to meet in practice. In this embodiment, the Copula model is used for statistical modeling in this scenario. The Copula model decomposes the multidimensional distribution into two parts, the marginal distribution and the dependency structure, through the Copula function, so as to flexibly model the dependency relationship between variables. Copula technology can realize the characterization of the nonlinear correlation of multivariate risk factors, and adopts layered Copula technology for bonds at all levels to prevent overfitting of single-layer Copula, so as to more accurately and scientifically realize complex stress scenario modeling, and provide possibilities for setting a wider range of risk scenarios.

[0041] Copula function is a multidimensional distribution function whose marginal distribution is uniform. According to Sklar's theorem, any multidimensional joint distribution can be expressed as a combination of its marginal distribution and a Copula function. Specifically, suppose there are n random variables X1, X2, ..., X n , their joint distribution function is F(X1, X2, ..., X n), the marginal distributions are F(X1), F(X2), ..., F(X n ), then there exists a Copula function C such that formula (1) satisfies:

[0042] F(X1, X2, ..., X n )=C(F(X1),F(X2),...,F(X n )) (1)

[0043] If F(X1), F(X2), ..., F(X n ) is continuous, C is uniquely determined. If C is a Copula function, F(X1), F(X2), ..., F(X n ) is a univariate distribution function, then the function F is a marginal distribution F(X1), F(X1), ..., F(X n ) is the joint distribution function of .

[0044] According to Sklar's theorem, more joint distribution functions can be constructed using marginal distribution functions and Copula functions, rather than just being limited to the assumptions of multivariate normal distribution or multivariate t distribution of data. There are many types of Copula functions, and common Copula functions include Gaussian Copula, t-Copula, Archimedean Copula, etc. Each type of Copula function has its own specific dependency structure characteristics, as shown in Table 2.

[0045] Table 2 Characteristics of different types of Copula functions

[0046]

[0047] It should be noted that the Copula model is a statistical model used to capture the complex dependencies between multiple random variables. It has the following advantages in the stress test scenario setting: First, it can accurately capture the linkage law between multiple yields. Bond yields often present an asymmetric distribution. For example, during the period when the bond market is poor, the yield fluctuation tends to be upward, which is an asymmetric skewed distribution. The Copula model can model this nonlinear and asymmetric correlation, which can more accurately reflect the linkage law of yields; second, it can reasonably characterize the distribution characteristics of bond yield samples. The Copula model is relatively flexible and can fit sample data with different distribution laws by setting model functions. The bond yield fluctuation series has the characteristics of "peaked and thick tail" and the tail correlation is strong. The t-Copula model can be used to fit this yield sample distribution characteristic; third, it can integrate the refinement advantages of the historical percentile method. Based on statistical models such as the Copula model, a large number of scenario samples can be generated. Sorting the yield fluctuations of benchmark bonds can obtain stress scenarios under different percentiles, providing users with refined choices.

[0048] In this embodiment, in order to facilitate model fitting, different bond yields can be layered and modeled according to the granularity, that is, according to the characteristics of bond yields, they are layered according to the granularity, and modeling is performed within each layer. For example, Figure 3 As shown in the figure, it can be divided into primary classification and secondary classification, where the primary classification is a rough classification, for example, it can be divided into interest-bearing bonds, financial bonds, and corporate bonds according to the bond type; the secondary classification is a detailed classification based on the primary classification, for example, interest-bearing bonds can be subdivided into treasury bonds, national development bonds, and local bonds. Copula models are used for modeling sampling at each level of classification, and finally spliced ​​into multiple bond volatility stress scenarios.

[0049] It should be noted that the bond types involved in the stress scenarios can be divided according to the focus of users (relevant personnel using the stress testing method provided in this implementation). For example, if the stress test focuses on interest-bearing bonds, interest-bearing bonds can be subdivided into treasury bonds, local bonds, policy-based financial bonds, etc., to provide a more refined stress scenario. For example, if the stress test focuses on financial bonds, financial bonds can be further subdivided as the next layer of bonds. The specific stratification method can be configured as needed, and this is not limited in this embodiment.

[0050] For each target bond, its historical yield fluctuation sequence is the yield fluctuation sequence determined by calculating the yield fluctuation according to the set sample frequency for the historical yield curve of the specified time span. Here, the time span determines the number of samples entering the model. A large sample size is conducive to improving the accuracy of modeling. For example, selecting the full amount of historical yield samples. However, if the sample time span is too long, there may be a problem of lack of timeliness of data, which will cause the model fitting results to deviate from the real market environment. Therefore, the time span of historical yields can be specified according to the needs of stress testing. The Copula model uses historical yield change values ​​as samples, and the sample frequency refers to the time span corresponding to the yield change. Since the shorter time span is often smaller than the yield change of the longer time span, the sample frequency selection will directly affect the result of the stress scenario setting. The sample frequency can be selected by referring to the implementation of stress test objectives and frequencies. For example, if you want to conduct a stress test every quarter to examine the ability of the target to bear risks in the next quarter, you can consider the corresponding stress scenario for the time span of the next quarter.

[0051] Each target bond may correspond to a distribution type, that is, to a marginal distribution function. The types of marginal distribution functions corresponding to different target bonds may be the same or different. The parameters of the marginal distribution function corresponding to each target bond may be determined by using the historical yield fluctuation sequence of each target bond. Here, the parameters of the marginal distribution function corresponding to each target bond may be determined by using the marginal distribution function corresponding to each target bond to fit the historical yield fluctuation sequence of each target bond, thereby ensuring that the fitted marginal distribution function can better characterize the historical yield fluctuation sequence of each target bond.

[0052] Based on the historical yield fluctuation series of the target bonds in each level, the model parameters of the Copula model corresponding to each level can be calculated, and based on the model parameters of the Copula model corresponding to each level and the parameters of the marginal distribution function corresponding to the target bonds in each level, the multidimensional joint distribution function corresponding to the target bonds in each level can be determined. The determination process can refer to the Sklar theorem, which will not be elaborated in this embodiment.

[0053] Based on the multidimensional joint distribution function corresponding to the target bonds in each level, a set of simulated number pairs of corresponding yield fluctuations can be generated for the target bonds in each level respectively. The simulated number pairs of corresponding yield fluctuations generated for each target bond in each level can be simulated number pairs that conform to the multidimensional joint distribution function corresponding to the target bonds in each level, or can be generated using number pairs that conform to the multidimensional joint distribution function corresponding to the target bonds in each level. This is not limited in this embodiment.

[0054] A simulated number pair generated for a level contains a simulated number (a yield fluctuation value) for each target bond in the level. Considering that multiple target bonds are used for stress testing of the target under test, the simulated number pairs generated for different levels can be spliced ​​to obtain the stress scenario of the target under test. The simulated number pairs generated for different levels used for splicing can be determined based on the association relationship between the associated bonds in different levels.

[0055] For example, the first-level bond 1 and the second-level bond 2 are related bond types, which can be the same bond type or related bond types (for example, bond type 2 is a subordinate bond type of 1). For the simulated number pairs generated for the first level, the simulated number pairs generated for the second level can be searched based on the simulated number 1 of bond type 1, and the simulated number pairs generated for the first level and the simulated number pairs generated for the second level found can be spliced, wherein, in the simulated number pairs found, the simulated number 2 of bond type 2 matches the simulated number 1 of bond type 1; or, the simulated number pairs generated for the second level can be spliced ​​based on the simulated number 1 of bond type 1, wherein, in the generated simulated number pairs, the simulated number 3 of bond type 2 matches the simulated number 1 of bond type 1.

[0056] It should be noted that each level contains at least two target securities, the number of target securities contained in each level can be the same or different, and one target securities can belong to multiple levels. The number of simulated number pairs contained in a set of simulated number pairs generated for each level can be the same or different.

[0057] After obtaining a set of stress scenarios, all stress scenarios can be used to perform stress tests on the target under test to obtain stress test results for the target under test. Alternatively, based on the user's choice or other conditions, some stress scenarios can be selected to perform stress tests on the target under test to obtain stress test results for the target under test. Using stress scenarios to perform stress tests on the target under test may include but is not limited to at least one of the following: collecting test data, setting assumptions, determining test methods, performing stress tests, and analyzing test results. The specific process can be referred to in the relevant technology and will not be described in detail in this embodiment.

[0058] Through the embodiments of the present application, the historical yield fluctuation sequence of each target bond among the multiple target bonds held by the measured target is determined, wherein the multiple target bonds are divided into multiple levels, each level in the multiple levels contains at least two target bonds among the multiple target bonds, and each level corresponds to a Copula model; the parameters of the marginal distribution function corresponding to each target bond are determined by using the historical yield fluctuation sequence of each target bond; based on the historical yield fluctuation sequence of the target bonds in each level, the model parameters of the Copula model corresponding to each level are calculated, and according to the model parameters of the Copula model corresponding to each level and the model parameters of the Copula model corresponding to each level, the model parameters of the Copula model corresponding to each level are calculated; The parameters of the marginal distribution function corresponding to the target bonds in the hierarchy are used to determine the multidimensional joint distribution function corresponding to the target bonds in each hierarchy; based on the multidimensional joint distribution function corresponding to the target bonds in each hierarchy, a set of simulated number pairs of corresponding yield fluctuations are generated for the target bonds in each hierarchy respectively, and the set of simulated number pairs generated for each hierarchy are spliced ​​to obtain a set of stress scenarios for the target under test; at least part of the stress scenarios in the set of stress scenarios are used to perform stress testing on the target under test to obtain stress test results of the target under test, which solves the problem of poor accuracy of test results in the stress testing method in the related art and improves the accuracy and rationality of the test results.

[0059] In an exemplary embodiment, the historical yield fluctuation sequence of each target bond among multiple target bonds held by the target under test is determined, including: screening out bonds that meet multiple targets from a preset bond set according to a set of measurement dimensions, and extracting the historical yield fluctuation sequence of each target bond.

[0060] In this embodiment, in order to achieve a more accurate, comprehensive and customized stress scenario, at the user level, a set of measurement dimensions of sample data can be set, and the yield fluctuations can be considered from multiple dimensions. The set set of measurement dimensions may include at least one of the following: time span, sample frequency, bond type, bond rating, and remaining term. Among them, the time span, sample frequency and bond type are similar to those in the aforementioned embodiment and are not repeated here. According to the set set of measurement dimensions, multiple target bonds can be screened out from the preset bond set, and the historical yield fluctuation sequence of each target bond can be extracted. Here, the method of extracting the historical yield fluctuation sequence of each target bond can be to extract the historical yield fluctuation sequence of each target bond from the pre-generated historical yield fluctuation sequence of candidate bonds, or, based on the historical yield of each target bond, generate the historical yield fluctuation sequence of each target bond.

[0061] For bond ratings, the fluctuation patterns of bond yield curves of bonds with different ratings under the same type are not the same. Users can further set bond ratings based on the bond type. For example, select AAA, AA+, AA, etc., and improve the accuracy of bond stress test scenarios through more detailed rating selection. The way to rate bonds can refer to multiple dimensions, and the set level types are also different, as long as the rating results can be guaranteed to be consistent with the actual situation in the consideration dimensions.

[0062] For the remaining maturity, the fluctuations of bonds with different remaining maturity in the stress scenario under extreme market conditions have their own rules, and there is a certain division between the long-term and short-term bond markets. Therefore, users can select the specific maturity of the bond type to obtain more accurate stress test results.

[0063] In addition, for the case where quantiles are used to set risk scenarios (i.e., stress scenarios), the set of measurement dimensions set at the user level can also include quantile levels. The set quantiles can be specific values ​​or percentiles based on benchmark values ​​and divided according to specified spans.

[0064] Exemplarily, the optional dimensions when setting a stress scenario (yield fluctuation) may be as shown in Table 3.

[0065] Table 3 Optional dimensions when setting pressure scenarios

[0066]

[0067] Here, the multi-dimensional settings in the stress test scenario can provide users with stress scenario setting conditions covering six dimensions: time span, sample frequency, bond type, bond rating, remaining term, and quantile level. Users can formulate customized, refined, and comprehensive stress test scenarios based on the characteristics of their own position data, and calculate the fair value fluctuation range based on the position institutions.

[0068] Through this embodiment, the flexibility of bond setting can be improved by screening target bonds based on at least one of time span, sample frequency, bond type, bond rating and remaining term.

[0069] In an exemplary embodiment, in multiple levels, the upper target bond containing multiple lower target bonds is one of the multiple lower target bonds, and each level contains a designated benchmark bond. For example, the first-level classification is interest-bearing bonds, financial bonds, and corporate bonds, and the second-level classification is the subdivision of interest-bearing bonds into treasury bonds, national development bonds, and local bonds. The interest-bearing bonds in the first-level classification are represented by treasury bonds, so the first-level classification is: treasury bonds, financial bonds, and corporate bonds, and the second-level classification is: treasury bonds, national development bonds, and local bonds. Designate treasury bonds as benchmark bonds, and both the first-level classification and the second-level classification contain designated benchmark bonds, that is, treasury bonds.

[0070] In this embodiment, based on the multidimensional joint distribution function corresponding to the target bonds in each level, a set of simulated number pairs of yield fluctuations are generated for the target bonds in each level, including: using Monte Carlo simulation to generate random number pairs of yield fluctuations that conform to the multidimensional joint distribution function corresponding to the target bonds in each level for each level, and obtain a set of random number pairs corresponding to each level, wherein a random number pair corresponding to each level includes a random number of the benchmark bond; based on the set of random number pairs corresponding to each level, a set of simulated number pairs is generated for each level.

[0071] Here, setting stress scenarios based on expert experience is too subjective, and setting stress scenarios based on historical data through a simple quantile method does not include predictions of future extreme scenarios. However, within each level, Monte Carlo simulation is used to obtain random pairs of yield fluctuations that conform to the joint distribution of various types of bonds. Monte Carlo simulation is applied to cover potential risk scenarios, so that the set stress scenarios can cover both historical extreme situations and predictions of future extreme scenarios, which is in line with the definition of stress testing itself.

[0072] Based on a set of random number pairs corresponding to each level, there can be multiple sets of simulated number pairs generated for each level, for example, a certain number of random numbers are selected as a set of simulated number pairs generated for each level. In order to improve the rationality of the simulated number pairs generated for each level, a more extensive risk scenario setting can be performed using a refined quantile level: based on each quantile in a set of quantiles of the benchmark bond volatility, a random number pair corresponding to each quantile is selected from a set of random number pairs corresponding to each level, wherein a set of quantiles is a quantile determined within a preset quantile range with a specified quantile as a span; based on the random number pairs corresponding to each quantile selected for each level, a simulated number pair corresponding to each quantile is generated for each level, and a set of simulated number pairs generated for each level is obtained.

[0073] Here, the preset quantile range can be a range starting from a certain quantile and spanning a certain quantile interval, for example, 80% to 90%, 80% to 95% or other ranges. The specified quantile can be a specified quantile span, for example, 0.5%, 1%, 1.5% or other values.

[0074] For example, when generating a large number of sample scenarios, the selected benchmark yield fluctuations are sorted and the percentiles are divided with a span of 0.5% to obtain stress scenarios with different quantile levels. Users do not need to specify the quantile level in advance, and can select the appropriate scenario in the output fine percentile scenario table.

[0075] Based on each quantile, a random number pair corresponding to each quantile can be screened out from a group of random number pairs corresponding to each level. The screening method can be: for each level, a certain target bond is used as the benchmark bond, and the simulation numbers of the benchmark bond are sorted in ascending order, and the random number pairs corresponding to each quantile are screened out from them. Based on the random number pairs corresponding to each quantile screened out at each level, simulated number pairs corresponding to each quantile are generated for each level. The method for generating the simulated number pairs can be: averaging the random numbers of the same target bond to obtain the simulated numbers at the corresponding positions.

[0076] For example, for the first-level classification, the corresponding random number pairs are 10,000. Taking interest-bearing bonds (e.g., government bonds) as the benchmark bonds, the random numbers (yield fluctuations) of interest-bearing bonds are sorted from small to large, and the quantile is 90%. Then, the random number pairs within a certain range around the 9000th random number are screened (e.g., with a span of 0.5, the 9001st to 9050th random numbers are screened); the random numbers of the screened random number pairs are averaged to obtain the simulated number pairs corresponding to the first-level classification and the 90% quantile.

[0077] The relevant schemes only give risk scenario settings for two to three key quantiles. Through this embodiment, a large number of stress scenario samples that obey the historical distribution law are generated by using the Copula model plus Monte Carlo numerical simulation technology. The generated samples can be used directly for stress testing, or users can select appropriate stress scenarios from them, or calculate the value at risk (VAR value) at different risk levels through a large number of stress scenarios, thereby improving the richness and rationality of stress scenario generation.

[0078] In an exemplary embodiment, in order to facilitate data association and improve the convenience of data processing, the benchmark bonds of each level can be the same; for different levels, the random number pairs corresponding to each quantile can be screened in the same or different ways.

[0079] For the first level among multiple levels, based on each quantile in a group of quantiles of the fluctuation of the benchmark bond, a random number pair corresponding to each quantile is screened out from a group of random number pairs corresponding to each level, including: from a group of random number pairs corresponding to the first level, a random number pair whose random number of the benchmark bond included is located in the quantile range corresponding to each quantile is screened out, and a plurality of random number pairs corresponding to each quantile screened out at the first level are obtained. The method of screening out random number pairs whose random number of the benchmark bond included is located in the quantile range corresponding to each quantile from a group of random number pairs corresponding to the first level is similar to that in the aforementioned embodiment, and will not be repeated here. Here, the quantile range corresponding to each quantile is a quantile range based on each quantile and with a specified quantile as a span. Taking the quantile of 80% and the span of 0.5 as an example, the quantile range corresponding to the 80% quantile is: 80% to 80.5%.

[0080] Correspondingly, based on the random number pairs corresponding to each quantile screened out at each level, simulated number pairs corresponding to each quantile are generated for each level, and a set of simulated number pairs generated for each level is obtained, including: taking the average of the multiple random number pairs corresponding to each quantile screened out at the first level, determining the simulated number pairs corresponding to each quantile generated at the first level, and obtaining a set of simulated number pairs generated for the first level. Here, the averaging method is to take the average of the random numbers of the same target bond respectively, then, a simulated number pair can be determined for a quantile, and a set of simulated number pairs generated for the first level corresponds one-to-one to a set of quantiles.

[0081] Through this embodiment, for the first level, random number pairs of a certain quantile level are screened according to the random numbers of the benchmark bonds, and corresponding simulated number pairs are generated by taking the average of the screened random number pairs, which can improve the convenience and rationality of the generation of simulated number pairs.

[0082] In an exemplary embodiment, for each quantile, each simulated number pair generated for the first level contains a simulated value of the yield fluctuation of the benchmark bond. For levels other than the first level, the same or similar method as described above can be used to obtain a set of simulated number pairs generated for other levels. Taking into account the correlation between benchmark bonds, when screening the random number pairs corresponding to each quantile from a set of random number pairs corresponding to other levels, the screening operation can be performed with reference to the simulated number pairs generated for the first level. For example, the screening operation can be performed with reference to the random number range corresponding to the benchmark bond and each quantile. The random number range corresponding to the benchmark bond and each quantile is based on the simulated number of the yield fluctuation of the benchmark bond in the simulated number pairs corresponding to each quantile of the first level, and the random number range with a specified value as the span. The span in this embodiment can be a span in one direction or a span in multiple directions, which is not limited in this embodiment.

[0083] For example, the simulated number of the yield fluctuation of the benchmark bond in the simulated number pair corresponding to a quantile at the first level is 90, and the specified value is 10%. The random number range corresponding to the benchmark bond and this quantile is: 90-99. If the span is an upper and lower span, the random number range corresponding to the benchmark bond and this quantile can be: 84-95, 85-94 or other ranges.

[0084] Correspondingly, for the other levels except the first level in the multiple levels, based on each quantile in a group of quantiles of the fluctuation of the benchmark bond, a random number pair corresponding to each quantile is selected from a group of random number pairs corresponding to each level, and further includes: selecting a random number pair whose random number of the benchmark bond is within the range of the random number corresponding to each quantile of the benchmark bond from a group of random number pairs corresponding to other levels, and obtaining a plurality of random number pairs corresponding to each quantile selected from other levels. Here, the number of random number pairs selected is not a fixed value, but is related to the distribution of the random number of the benchmark bond in the generated random number pairs.

[0085] For the random number pairs corresponding to each quantile screened out from other levels, based on the random number pairs corresponding to each quantile screened out from each level, simulated number pairs corresponding to each quantile are generated for each level, and a group of simulated number pairs generated for each level are obtained, which also includes: taking the average of the multiple random number pairs corresponding to each quantile screened out from other levels, determining the simulated number pairs corresponding to each quantile generated from other levels, and obtaining a group of simulated number pairs generated for other levels. The way of taking the average of the multiple random number pairs corresponding to each quantile screened out from other levels is similar to that in the aforementioned embodiment, and will not be repeated here. Similarly, a group of simulated number pairs generated for other levels corresponds one-to-one to a group of quantiles.

[0086] It should be noted that, in the present embodiment, different methods are used to screen the random number pairs corresponding to each quantile for the first level and other levels except the first level. If there is no contradiction, the random number pairs can be screened for the specified level (the specified level can be the first level or non-first level) in the manner of screening the random number pairs corresponding to each quantile for the first level as described above, and the random number pairs can be screened for other levels except the specified level in the manner of screening the random number pairs corresponding to each quantile for other levels as described above. The processing process is similar and will not be elaborated here.

[0087] Through this embodiment, the random number pair screening operation is performed on the random number range corresponding to each quantile of the reference benchmark bonds of other levels except the first level, which can improve the rationality of the random number pair screening.

[0088] In an exemplary embodiment, a set of analog number pairs is generated for each level in a manner similar to that in the aforementioned embodiment, and a set of analog number pairs generated for each level corresponds one-to-one to a set of bit numbers.

[0089] In this embodiment, the analog number pairs can be spliced ​​according to the quantiles. Correspondingly, a set of analog number pairs generated for each level are spliced ​​to obtain a set of pressure scenarios of the measured target, including: taking each quantile in a set of quantiles as the current quantile and performing the following splicing operation to obtain the pressure scenario corresponding to the measured target and each quantile: splicing the analog number pairs generated for each level and corresponding to the current quantile to obtain the pressure scenario corresponding to the measured target and the current quantile.

[0090] Since the same target bonds are included in multiple levels, for example, the aforementioned benchmark bonds, when the simulated number pairs generated for each level and corresponding to the current quantile are spliced, multiple simulated numbers of the benchmark bonds may be included, and the multiple simulated numbers of the benchmark bonds may not be the same. To this end, the simulated numbers of the benchmark bonds of the specified level (first level or non-first level) can be used as the standard simulated numbers in a geometric adjustment manner, and the simulated values ​​of other bonds in other levels except the benchmark bonds can be adjusted according to the ratio of the simulated numbers of the benchmark bonds of other levels to the standard simulated numbers, and only one simulated value of the benchmark bond can be retained in the stress scenario after splicing.

[0091] Optionally, in order to improve the convenience of stress scenario generation, the simulation numbers of benchmark bonds of other levels other than the specified level can be deleted, and the remaining simulation values ​​can be concatenated. Correspondingly, the simulation number of the benchmark bond in the stress scenario corresponding to the measured target and the current quantile is the simulation number of the benchmark bond in the simulation number pair corresponding to the current quantile generated for the specified level in the multiple levels.

[0092] Through this embodiment, simulated number pairs are concatenated according to quantiles, and the simulated values ​​of the benchmark bonds retain the simulated values ​​of the benchmark bonds at the specified level, which can improve the convenience of stress scenario generation.

[0093] In an exemplary embodiment, when using the historical yield fluctuation sequence of each target bond to determine the parameters of the marginal distribution function corresponding to each target bond, each target bond can be used as the current target bond to perform the following determination operations to obtain the parameters of the marginal distribution function corresponding to each target bond: use each marginal distribution function in a set of marginal distribution functions to fit the historical yield fluctuation sequence of the current target bond, and obtain the fitting results corresponding to each marginal distribution function; determine the marginal distribution function in a set of marginal distribution functions whose corresponding fitting result has the highest match with the historical yield fluctuation sequence of the current target bond as the marginal distribution function corresponding to the current target bond, and obtain the current marginal distribution function.

[0094] For the current target bond, each marginal distribution function in a set of marginal distribution functions can be used to fit the historical yield fluctuation sequence of the current target bond to obtain the fitting result corresponding to each marginal distribution function. The fitting result here includes the parameters of each fitted marginal distribution function. The fitting results corresponding to different marginal distribution functions have different matching degrees with the historical yield fluctuation sequence of the current target bond. The marginal distribution function in a set of marginal distribution functions whose corresponding fitting result matches the historical yield fluctuation sequence of the current target bond the most can be determined as the marginal distribution function corresponding to the current target bond to obtain the current marginal distribution function. Correspondingly, the parameters of the current marginal distribution function are the parameters of the current marginal distribution function in the fitting result corresponding to the current marginal distribution function.

[0095] In addition, when the matching degree between the fitting result corresponding to a certain marginal distribution function and the historical yield fluctuation sequence of the current target bond reaches the set matching degree threshold, other marginal distribution functions will no longer be judged. In this case, in order to facilitate the determination of the corresponding marginal distribution function, the marginal distribution function corresponding to the current target bond can be determined by referring to the target bond whose corresponding marginal distribution function has been determined. For example, the marginal distribution function corresponding to the target bond with the same sample frequency can be used as the marginal distribution function with priority matching, and for another example, the marginal distribution function corresponding to the target bond at the same level can be used as the marginal distribution function with priority matching.

[0096] Through this embodiment, according to the matching degree between the fitting results corresponding to different marginal distribution functions and the historical yield fluctuation sequence of the current target bond, the marginal distribution function corresponding to the current target bond is selected, so as to improve the accuracy of determining the marginal distribution function.

[0097] The stress testing method in the embodiment of the present application is described below in conjunction with an optional example. This optional example provides a method for setting a Copula model for a stress scenario, which can be performed by a terminal device or a terminal device in combination with a server, and includes user-level and system-level operations, wherein the user level reflects the operations performed by the user on the terminal device or the results that the user can view on the terminal device, and the system level is the operations performed by the system running on the terminal device or the background server for setting the stress scenario.

[0098] Figure 4 is a flow chart of another optional stress testing method according to an embodiment of the present application, such as Figure 4 As shown, the process of the method mainly includes steps S402 to S410.

[0099] Step S402, determine the sample. The user sets the sample data measurement dimension on the terminal device, which may include: time span, sample frequency, bond type, bond rating, remaining term and quantile level, etc. The terminal device then extracts the historical fluctuation series of various bond yields of interest and stratifies them according to the granularity, thereby setting the model level.

[0100] Step S404, determining marginal distribution: Determine the marginal distribution type of the fluctuation of the yield of each type of bond, and estimate the parameters of the marginal distribution function of each type of bond using historical data.

[0101] Step S406, determine the joint distribution. Determine the copula type of the joint distribution of the volatility of the yields of various types of bonds in each layer (ie, level), and calculate the parameters of the copula of this type using historical data, so as to obtain the parameters of the joint distribution function.

[0102] Step S408, generating simulation values. In each layer, a random number pair of yield fluctuations that conforms to the joint distribution of various bonds is obtained according to the Monte Carlo simulation.

[0103] Step S410, determine the stress scenario. In each layer, the corresponding stress scenario of each level is obtained according to the quantile level of the benchmark bond fluctuation, and the layered data is spliced ​​to generate a visualization result, and the above visualization result is displayed to the user on the terminal device.

[0104] For example, the stress test scenario settings for three types of bonds, namely government bonds, ordinary bank bonds and AAA-rated corporate bonds, are used as an example. The sample time span is selected from the specified year to the present, the frequency is selected from daily frequency, weekly frequency, monthly frequency, quarterly frequency, semi-annual frequency and annual frequency, and the bond types are selected from government bonds, ordinary bank bonds and AAA-rated corporate bonds. The time series of the three types of bonds are calculated according to the mean of the six key terms of 1Y, 2Y, 3Y, 5Y, 7Y and 10Y. Since the yield fluctuation sequence has the characteristics of "peaked thick tail", the selection of t distribution can better fit the data of the three types of bonds. The Copula model selects t-Copula to characterize the tail correlation between financial time series variables.

[0105] Taking the national debt as the benchmark, the average value of the increase and decrease of the random number pairs of the three types of bonds at different percentiles is calculated to obtain the output results. Table 4 shows the key percentile results of the stress test scenario settings of the first-level bonds at different frequencies.

[0106] Table 4 Example table of results of different frequency yield data settings (unit: BP)

[0107]

[0108] Table 5 shows the stress test scenarios obtained by applying multi-layer Copula modeling to the secondary bonds with monthly yields, showing example results for the 80%, 85%, 90%, 95%, and 99% quantiles.

[0109] Table 5 Example of monthly yield rate data setting results (unit: BP)

[0110]

[0111] Through this optional example, the Copula model is used to capture the complex dependencies between multiple bond yields, which can accurately capture the linkage patterns between multiple yields and reasonably characterize the distribution characteristics of bond yield samples. A large number of scenario samples can be generated based on the Copula model, and the yield fluctuations of the benchmark bonds can be ranked to obtain stress scenarios at different percentiles, providing users with refined choices and improving the rationality and convenience of stress scenario generation.

[0112] Through the description of the above implementation methods, those skilled in the art can clearly understand that the method according to the above embodiment can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods described in each embodiment of the present application.

[0113] According to another aspect of the embodiments of the present application, a pressure testing device is also provided, which is used to implement the pressure testing method provided in the above embodiments, and will not be repeated hereafter. As used below, the term "module" can implement a combination of software and / or hardware for a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceivable.

[0114] Figure 5 is a structural block diagram of an optional pressure testing device according to an embodiment of the present application, such as Figure 5 As shown in , the apparatus includes: a first determining unit 502 , a second determining unit 504 , a first executing unit 506 , a second executing unit 508 and a testing unit 510 .

[0115] The first determination unit 502 is used to determine the historical yield fluctuation sequence of each target bond among the multiple target bonds held by the target under test, wherein the multiple target bonds are divided into multiple levels, each level in the multiple levels contains at least two target bonds among the multiple target bonds, and each level corresponds to a Copula model. The first determination unit 502 can be used to execute step S202.

[0116] The second determination unit 504 is used to determine the parameters of the marginal distribution function corresponding to each target bond by using the historical yield fluctuation sequence of each target bond. The second determination unit 504 can be used to execute step S204.

[0117] The first execution unit 506 is used to calculate the model parameters of the Copula model corresponding to each level based on the historical yield fluctuation sequence of the target bond in each level, and determine the multidimensional joint distribution function corresponding to the target bond in each level according to the model parameters of the Copula model corresponding to each level and the parameters of the marginal distribution function corresponding to the target bond in each level. The first execution unit 506 can be used to execute step S206.

[0118] The second execution unit 508 is used to generate a set of simulated number pairs of yield fluctuations corresponding to the target bonds in each level based on the multidimensional joint distribution function corresponding to the target bonds in each level, and to splice the set of simulated number pairs generated for each level to obtain a set of stress scenarios for the measured target. The second execution unit 508 can be used to execute step S208.

[0119] The testing unit 510 is used to perform a stress test on the target under test using at least part of the stress scenarios in a set of stress scenarios to obtain a stress test result of the target under test. The testing unit 510 can be used to execute step S210.

[0120] Through the embodiments of the present application, the historical yield fluctuation sequence of each target bond among the multiple target bonds held by the measured target is determined, wherein the multiple target bonds are divided into multiple levels, each level in the multiple levels contains at least two target bonds among the multiple target bonds, and each level corresponds to a Copula model; the parameters of the marginal distribution function corresponding to each target bond are determined by using the historical yield fluctuation sequence of each target bond; based on the historical yield fluctuation sequence of the target bonds in each level, the model parameters of the Copula model corresponding to each level are calculated, and according to the model parameters of the Copula model corresponding to each level and the model parameters of the Copula model corresponding to each level, the model parameters of the Copula model corresponding to each level are calculated; The parameters of the marginal distribution function corresponding to the target bonds in the hierarchy are used to determine the multidimensional joint distribution function corresponding to the target bonds in each hierarchy; based on the multidimensional joint distribution function corresponding to the target bonds in each hierarchy, a set of simulated number pairs of corresponding yield fluctuations are generated for the target bonds in each hierarchy respectively, and the set of simulated number pairs generated for each hierarchy are spliced ​​to obtain a set of stress scenarios for the target under test; at least part of the stress scenarios in the set of stress scenarios are used to perform stress testing on the target under test to obtain stress test results of the target under test, which solves the problem of poor accuracy of test results in the stress testing method in the related art and improves the accuracy and rationality of the test results.

[0121] In an exemplary embodiment, in multiple levels, the upper target bond containing multiple lower target bonds is one of the multiple lower target bonds, and each level contains a specified benchmark bond. The second execution unit includes: a first generation module, which is used to use Monte Carlo simulation to generate random number pairs corresponding to the yield fluctuations of the multidimensional joint distribution function corresponding to the target bonds in each level for each level, and obtain a group of random number pairs corresponding to each level, wherein a random number pair corresponding to each level contains a random number of the benchmark bond; a screening module, which is used to screen out the random number pairs corresponding to each quantile from the group of random number pairs corresponding to each level based on each quantile in a group of quantiles of the benchmark bond fluctuations, wherein a group of quantiles is a quantile determined within a preset quantile range with a specified quantile as a span; a second generation module, which is used to generate simulation number pairs corresponding to each quantile for each level based on the random number pairs corresponding to each quantile screened out for each level, and obtain a group of simulation number pairs generated for each level.

[0122] In an exemplary embodiment, the screening module includes: a first screening submodule, for screening, for a first level among multiple levels, random number pairs whose random numbers of the benchmark bonds are included in a set of random number pairs corresponding to the first level, and obtaining multiple random number pairs corresponding to each quantile screened out at the first level, wherein the quantile range corresponding to each quantile is a quantile range based on each quantile and spanning a specified quantile. The second generation module includes: a first generation submodule, for determining the result of taking the average of the multiple random number pairs corresponding to each quantile screened out at the first level as the simulated number pairs generated at the first level and corresponding to each quantile, and obtaining a set of simulated number pairs generated for the first level, wherein the set of simulated number pairs generated for the first level corresponds one-to-one to a set of quantiles.

[0123] In an exemplary embodiment, each simulated number pair generated for the first level includes a simulated value of the yield fluctuation of the benchmark bond. The screening module also includes: a second screening submodule, which is used to screen out random number pairs containing the random number of the benchmark bond in a set of random number pairs corresponding to the other levels, for the other levels except the first level in the multiple levels, and obtain multiple random number pairs corresponding to each quantile screened out by the other levels, wherein the random number range corresponding to the benchmark bond and each quantile is a random number range based on the simulated number of the yield fluctuation of the benchmark bond in the simulated number pairs corresponding to each quantile of the first level, and with a specified value as the span. The second generation module also includes: a second generation submodule, which is used to average the results of the multiple random number pairs corresponding to each quantile screened out by the other levels, and determine them as the simulated number pairs corresponding to each quantile generated by the other levels, and obtain a set of simulated number pairs generated for the other levels, wherein the set of simulated number pairs generated for the other levels corresponds to a set of quantiles one-to-one.

[0124] In an exemplary embodiment, a set of simulated number pairs generated for each level corresponds to a set of quantile levels. The second execution unit includes: a first execution module, which is used to perform the following splicing operation on each quantile in a set of quantiles as the current quantile to obtain a stress scenario corresponding to the measured target and each quantile: splicing the simulated number pairs generated for each level and corresponding to the current quantile to obtain a stress scenario corresponding to the measured target and the current quantile, wherein the simulated number of the benchmark bond in the stress scenario corresponding to the measured target and the current quantile is the simulated number of the benchmark bond in the simulated number pair corresponding to the current quantile generated for a specified level in the multiple levels.

[0125] In an exemplary embodiment, the first determination unit includes: a second execution module, which is used to screen out bonds that meet multiple targets from a preset bond set according to a set of measurement dimensions, and extract the historical yield fluctuation series of each target bond, wherein the set of measurement dimensions includes at least one of the following: time span, sample frequency, bond type, bond rating, and remaining term.

[0126] In an exemplary embodiment, the second determination unit includes: a third execution module, which is used to perform the following determination operations on each target bond as the current target bond to obtain the parameters of the marginal distribution function corresponding to each target bond: use each marginal distribution function in a set of marginal distribution functions to fit the historical yield fluctuation sequence of the current target bond to obtain the fitting results corresponding to each marginal distribution function; determine the marginal distribution function in a set of marginal distribution functions whose corresponding fitting results have the highest match with the historical yield fluctuation sequence of the current target bond as the marginal distribution function corresponding to the current target bond to obtain the current marginal distribution function; wherein the parameters of the current marginal distribution function are the parameters of the current marginal distribution function in the fitting results corresponding to the current marginal distribution function.

[0127] It should be noted that the above modules can be implemented by software or hardware. For the latter, it can be implemented in the following ways, but not limited to: the above modules are all located in the same processor; or the above modules are located in different processors in any combination.

[0128] According to another aspect of the embodiments of the present application, a computer-readable storage medium is provided, wherein the computer-readable storage medium includes a stored program, wherein the steps of any of the above method embodiments are executed when the program is run.

[0129] In an exemplary embodiment, the computer-readable storage medium may include, but is not limited to, various media capable of storing computer programs, such as a USB flash drive, a ROM, a RAM, a mobile hard disk, a magnetic disk, or an optical disk.

[0130] According to another aspect of the embodiments of the present application, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor is configured to execute the steps in any of the above method embodiments through the computer program. In an exemplary embodiment, the electronic device may further include a transmission device and an input / output device, wherein the transmission device is connected to the processor, and the input / output device is connected to the processor.

[0131] For specific examples in this embodiment, reference may be made to the examples described in the above embodiments and exemplary implementation modes, and this embodiment will not be described in detail herein.

[0132] According to another aspect of the embodiment of the present application, a computer program product is also provided, which includes a computer program / instruction, and the computer program / instruction contains a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 609, and / or installed from the removable medium 611. When the computer program is executed by the central processing unit 601, various functions provided by the embodiment of the present application are executed. The above-mentioned serial numbers of the embodiments of the present application are only for description and do not represent the advantages and disadvantages of the embodiments.

[0133] Figure 6 The computer system structure block diagram of the electronic device used to implement the embodiment of the present application is schematically shown. Figure 6 As shown, the computer system 600 includes a CPU (Central Processing Unit) 601, which can perform various appropriate actions and processes according to the program stored in the ROM 602 or the program loaded from the storage part 608 to the RAM 603. Various programs and data required for system operation are also stored in the random access memory 603. The central processing unit 601, the read-only memory 602 and the random access memory 603 are connected to each other through a bus 604. An I / O (Input / Output) interface 605 is also connected to the bus 604.

[0134] The following components are connected to the I / O interface 605: an input section 606 including a keyboard, a mouse, etc.; an output section 607 including a CRT (Cathode Ray Tube), an LCD (Liquid Crystal Display), etc., and a speaker, etc.; a storage section 608 including a hard disk, etc.; and a communication section 609 including a network interface card such as a LAN card, a modem, etc. The communication section 609 performs communication processing via a network such as the Internet. A drive 610 is also connected to the input / output interface 605 as needed. A removable medium 611, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 610 as needed so that a computer program read therefrom is installed into the storage section 608 as needed.

[0135] In particular, according to an embodiment of the present application, the process described in each method flow chart can be implemented as a computer software program. For example, an embodiment of the present application includes a computer program product, which includes a computer program carried on a computer readable medium, and the computer program contains a program code for executing the method shown in the flow chart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 609, and / or installed from the removable medium 611. When the computer program is executed by the central processor 601, various functions defined in the system of the present application are executed.

[0136] It should be noted that Figure 6 The computer system 600 of the electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0137] Obviously, those skilled in the art should understand that the above modules or steps of the present application can be implemented by a general computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, and in some cases, the steps shown or described can be executed in a different order from that herein, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the present application is not limited to any specific combination of hardware and software.

[0138] The above are only preferred embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the principles of the present application shall be included in the protection scope of the present application.

Claims

1. A stress testing method, characterized in that: include: Determine the historical yield fluctuation sequence of each target bond among a plurality of target bonds held by the measured target, wherein the plurality of target bonds are divided into a plurality of levels, each level of the plurality of levels contains at least two target bonds among the plurality of target bonds, and each level corresponds to a Copula model; Determine the parameters of the marginal distribution function corresponding to each target bond by using the historical yield fluctuation sequence of each target bond; Based on the historical yield fluctuation series of the target bonds in each level, the model parameters of the Copula model corresponding to each level are calculated, and according to the model parameters of the Copula model corresponding to each level and the parameters of the marginal distribution function corresponding to the target bonds in each level, the multidimensional joint distribution function corresponding to the target bonds in each level is determined; Based on the multidimensional joint distribution function corresponding to the target bonds in each level, a set of simulated number pairs of corresponding yield fluctuations are generated for the target bonds in each level respectively, and the set of simulated number pairs generated for each level are spliced ​​to obtain a set of stress scenarios for the measured target; Perform a stress test on the target under test using at least some of the stress scenarios in the set of stress scenarios to obtain a stress test result of the target under test.

2. The method according to claim 1, characterized in that: In the plurality of levels, an upper level target bond including a plurality of lower level target bonds is one of the plurality of lower level target bonds, and each of the levels includes a designated benchmark bond; The method of generating a set of simulated number pairs of corresponding yield fluctuations for the target bonds in each level based on the multidimensional joint distribution function corresponding to the target bonds in each level comprises: Using Monte Carlo simulation, generating random number pairs of yield fluctuations that conform to the multidimensional joint distribution function corresponding to the target bonds in each level for each level, and obtaining a set of random number pairs corresponding to each level, wherein a random number pair corresponding to each level includes a random number of the benchmark bond; Based on each quantile in a group of quantiles of the volatility of the benchmark bond, a random number pair corresponding to each quantile is selected from a group of random number pairs corresponding to each level, wherein the group of quantiles is a quantile determined within a preset quantile range with a specified quantile as a span; Based on the random number pairs corresponding to each quantile screened out at each level, simulated number pairs corresponding to each quantile are generated for each level to obtain a group of simulated number pairs generated for each level.

3. The method according to claim 2, characterized in that The step of selecting, for each quantile in a group of quantiles based on the fluctuation of the benchmark bond, a random number pair corresponding to each quantile from a group of random number pairs corresponding to each level, comprises: for a first level among the multiple levels, selecting, from a group of random number pairs corresponding to the first level, a random number pair whose random number of the benchmark bond is within a quantile range corresponding to each quantile, to obtain a plurality of random number pairs corresponding to each quantile selected by the first level, wherein the quantile range corresponding to each quantile is a quantile range based on each quantile and spanning the specified quantile; The random number pairs corresponding to each quantile screened out from each level, generating simulated number pairs corresponding to each quantile for each level, and obtaining a group of simulated number pairs generated for each level, include: taking the average of multiple random number pairs corresponding to each quantile screened out from the first level, determining the simulated number pairs generated for the first level and corresponding to each quantile, and obtaining a group of simulated number pairs generated for the first level, wherein the group of simulated number pairs generated for the first level corresponds one-to-one to the group of quantiles.

4. The method according to claim 3, characterized in that Each simulated number pair generated for the first level comprises a simulated value of the yield volatility of the benchmark bond; The step of selecting a random number pair corresponding to each quantile in a group of quantiles based on the fluctuation of the benchmark bond from a group of random number pairs corresponding to each level also includes: for other levels except the first level in the multiple levels, selecting a random number pair whose random number of the benchmark bond is within the range of random numbers corresponding to the benchmark bond and each quantile from a group of random number pairs corresponding to the other levels, and obtaining a plurality of random number pairs corresponding to each quantile selected by the other levels, wherein the range of random numbers corresponding to the benchmark bond and each quantile is a random number range based on the simulated number of the yield fluctuation of the benchmark bond in the simulated number pairs corresponding to the first level and each quantile, and with a specified value as the span; The random number pairs corresponding to each quantile screened out from each level, generating simulated number pairs corresponding to each quantile for each level, and obtaining a group of simulated number pairs generated for each level, also includes: averaging the multiple random number pairs corresponding to each quantile screened out from the other levels, determining the simulated number pairs corresponding to each quantile generated for the other levels, and obtaining a group of simulated number pairs generated for the other levels, wherein the group of simulated number pairs generated for the other levels corresponds one-to-one to the group of quantiles.

5. The method according to claim 2, characterized in that: A set of simulated number pairs generated for each level corresponds one-to-one to the set of quantiles; The step of splicing a set of simulated number pairs generated for each level to obtain a set of pressure scenarios for the measured target includes: Each quantile in the group of quantiles is used as the current quantile to perform the following splicing operation to obtain the pressure scenario corresponding to the measured target and each quantile: The simulated number pairs generated for each level and corresponding to the current quantile are concatenated to obtain a stress scenario corresponding to the measured target and the current quantile, wherein the simulated number of the benchmark bond in the stress scenario corresponding to the measured target and the current quantile is the simulated number of the benchmark bond in the simulated number pair corresponding to the current quantile generated for a specified level among the multiple levels.

6. The method according to any one of claims 1 to 5, characterized in that The step of determining the historical yield fluctuation sequence of each target bond among the plurality of target bonds of the measured target comprises: According to a set of measurement dimensions, screen out the bonds that meet the multiple targets from a preset bond set, and extract the historical yield fluctuation series of each target bond, wherein the set of measurement dimensions includes at least one of the following: time span, sample frequency, bond type, bond rating, and remaining term.

7. The method according to any one of claims 1 to 5, characterized in that The method of using the historical yield fluctuation sequence of each target bond to determine the parameters of the marginal distribution function corresponding to each target bond includes: The following determination operations are performed on each target bond as the current target bond to obtain the parameters of the marginal distribution function corresponding to each target bond: Use each marginal distribution function in a set of marginal distribution functions to fit the historical yield fluctuation series of the current target bond, and obtain a fitting result corresponding to each marginal distribution function; Determine the marginal distribution function of the set of marginal distribution functions, whose corresponding fitting result matches the historical yield fluctuation sequence of the current target bond with the highest degree, as the marginal distribution function corresponding to the current target bond, to obtain the current marginal distribution function; The parameters of the current marginal distribution function are the parameters of the current marginal distribution function in the fitting result corresponding to the current marginal distribution function.

8. A pressure testing device, characterized in that: include: A first determining unit is used to determine a historical yield fluctuation sequence of each target bond among a plurality of target bonds held by the target under test, wherein the plurality of target bonds are divided into a plurality of levels, each level of the plurality of levels contains at least two target bonds among the plurality of target bonds, and each level corresponds to a Copula model; A second determination unit is used to determine the parameters of the marginal distribution function corresponding to each target bond by using the historical yield fluctuation sequence of each target bond; A first execution unit is used to calculate the model parameters of the Copula model corresponding to each level based on the historical yield fluctuation sequence of the target bonds in each level, and determine the multidimensional joint distribution function corresponding to the target bonds in each level according to the model parameters of the Copula model corresponding to each level and the parameters of the marginal distribution function corresponding to the target bonds in each level; A second execution unit is used to generate a set of simulated number pairs of yield fluctuations corresponding to the target bonds in each level based on the multidimensional joint distribution function corresponding to the target bonds in each level, and to splice the set of simulated number pairs generated for each level to obtain a set of stress scenarios for the measured target; The testing unit is used to perform a stress test on the target under test using at least part of the stress scenarios in the set of stress scenarios to obtain a stress test result of the target under test.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, wherein the computer program implements the steps of the method described in any one of claims 1 to 7 when executed by a processor.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method described in any one of claims 1 to 7 are implemented.

Images